The usefulness of the cross prediction method developed by J. L. Jinks and H. S. Pooni (1976, Heredity 36: 253–266) in a doubled-haploid breeding program was evaluated. Doubled-haploid lines were derived from two barley crosses by the bulbosum method. Doubled-haploid lines from each cross were divided into two sets, one with 52 lines and the other 58 to 79 lines. These doubled-haploid lines were evaluated for grain yield, heading date, and plant height in hill plots with 10 replications at Elora, Ontario, in 1973 and 1974. The mean and phenotypic variance of doubled-haploid lines from the first set were used to calculate the predicted distribution under the assumption of normality. The predicted distribution was compared with that observed from the other set by a χ2 test. It was found that the predicted distribution differed from the observed distribution in the majority of the cases and the two tails of the distribution accounted for much of the difference in 6 of the 12 (= 3 characters × 2 crosses × 2 years) cases. The numbers of transgressive lines and best lines were also predicted for the three characters in the two crosses in each year. Twenty of the 24 observed numbers of transgressive lines were not within their 95% confidence interval but 8 of the 12 observed numbers of best lines were within their 95% confidence interval. A significant correlation coefficient between predicted and observed numbers of transgressive lines was obtained for grain yield and heading date. The prediction that one cross had more high-yielding lines and fewer short-stature lines than the other cross agreed well with the actual observation. It appears that the cross prediction method is of value in predicting the cross potential and can be used for identifying superior crosses in a doubled-haploid breeding program in barley.Key words: barley, haploids, cross prediction.
